frank ramsey's notes on saving and taxation · 2019-03-11 · frank ramsey’s notes on saving...
Post on 21-Feb-2020
3 Views
Preview:
TRANSCRIPT
Frank Ramsey’s Notes on
Saving and Taxation
Edited by Pedro Garcia Duarte
Editor’s Introduction
There are two sets of undated and until now unpublished notes by Ram-
sey that support the idea that both of his papers published in the late
1920s, which are the predecessors of modern discussions on economic
growth, optimal taxation, and monetary economics, were part of a com-
mon research agenda, as discussed in the preceding article. These notes
are deposited at the Archives of Scientifi c Philosophy of the University
of Pittsburgh, Hillman Library (Papers of Frank Plumpton Ramsey,
series “Undergraduate Materials,” box 6, folder 7, “Ramsey Econom-
ics”) and are published here for the fi rst time by permission of that uni-
versity (all rights reserved).
The fi rst set of notes contains two sections (numbered IV and V) that
most likely were the ones cut out by Keynes from the published version
of Ramsey’s 1928 paper that is divided into three sections.1 In the next
section I reprint these notes, leaving out most of the passages struck out by
Ramsey in his manuscript.2 I sometimes include additional notes explain-
ing some passages, mostly as footnotes to Ramsey’s text. Lastly, in a few
instances Ramsey underlined words, which are here italicized.
History of Political Economy 41:3 DOI 10.1215/00182702-2009-048
Copyright 2009 by Duke University Press
1. The pages of these notes are numbered from 31 to 40. According to the classifi cation
system of the University of Pittsburgh, this set of annotations is numbered 006-07-01.
2. At a few places I do include, in double parentheses (( )), the text struck out by Ramsey.
472 History of Political Economy 41:3 (2009)
In these notes Ramsey modifi es his equations to consider the case of a
tax on savings. In section IV he explores the distinction between exhaus-
tive and transfer expenditures, which is, as argued in the preceding essay
in this issue, in line with Arthur Cecil Pigou’s analysis in his 1928 book on
public fi nance (chap. 3). He also considers, in the simple case present in
the published article of a linear production function (i.e., the case of con-
stant and independent returns to capital and labor), different tax rates on
earned and unearned incomes. Another interesting point is that these notes
constitute additional material to important discussions about Ramsey’s
enterprise in economics. For instance, with respect to the issue of whether
or not Ramsey had a notion of a representative agent, as we now under-
stand it, here we see Ramsey referring to the utility function “of an eternal
community” rather than of an individual (see Duarte 2008).
In section V of the manuscript we see Ramsey again following the dis-
cussion in Pigou’s 1928 book about the extent to which “a uniform income
tax should be remitted on savings.” It was in this context that Pigou was
one of the fi rst to refer to Ramsey’s 1927 result that lower tax rates should
be applied to the uses of income for which the demand is relatively elas-
tic (Duarte 2008). In the manuscripts Ramsey disagrees with Pigou’s use
of his result and Pigou’s conclusion that “as the demand for savings is
more elastic than that for consumption . . . there should be differentiation
in favour of savings.” As Ramsey makes clear, he does not “dispute the
soundness of this conclusion,” but he believes that his result cannot be
used to support it (see the full argument below). Ramsey then plays with
his equations in order to fi nd an answer to Pigou’s query. The mathematics
gets “extremely complicated” and Ramsey then considers the special case
of a fi xed labor supply, discussing in detail the several hypotheses made in
order to derive the set of equations in this manuscript.
The second set of notes are derivations of equations, some of which
Ramsey used in the fi rst set. I tried to add to these equations the side notes
that Ramsey wrote on some of his manuscript’s pages. At some places,
indicated in footnotes, his handwriting is not completely clear. I should
also warn readers that it is possible that I have made errors in editing some
of his equations, as some are not easy to read. The last pages of his manu-
script are particularly unclear and not fully developed. Here I indicate
these passages in footnotes.
For instance, in Ramsey’s annotations he seems to make a distinction
between the return to capital, a cursive r, and the interest rate used as dis-
count factor, a printed r. However, in a few places he seems to use both
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 473
3. Editor’s note: From the published article, equation (2) is v(a) = ∂∂—
f
au(x), where a(t) and x(t)
denote the total rates of labor and consumption at time t, v(a) = —dV
da
(—a) is the marginal disutility
of labor, and f(a, c) is the production function, with c(t) being the stock of capital at time t.
Equation (3) is —d
dtu(x(t)) = –
∂∂—
f
c ⋅ u(x(t)), where u(x(t)) = d—U
dx
(x—(t)
(t—)) is the marginal utility of con-
sumption. Using modern terminology, equation (2) gives us the intratemporal relationship
between labor and consumption while (3) represents the optimality condition for intertempo-
ral allocation of consumption. It is important to mention also that both equations refer to the
case of no discounting. Later in the paper, Ramsey (1928, 553–54) shows that when time-
discounting is introduced (again, with no taxes), equation (2) is unaltered, but (3) becomes
—d
dtu(x(t)) = –u(x(t)){
∂∂—
f
c – ρ}.
interchangeably and therefore I make no such distinction in the edited notes.
Lastly, in both sets of notes I kept Ramsey’s numbering of equations.
In the second set of notes we can see Ramsey crafting his mathematical
analysis on exempting savings from income taxes. He considers many spe-
cial cases, some discussed in the fi rst set of notes, others not. In some places
Ramsey even considers particular functional forms for the utility function,
which is something that he marginally used in the printed article for the
case of no taxation (p. 555). These are notes that are hard reading by them-
selves. Instead they should be explored together with the published arti-
cle and the fi rst set of notes. When compared with this fi rst set, they make
clear Ramsey’s compromise between mathematical derivations and sound
economic reasoning obtained from them, as Ramsey wrote to Harrod on
27 March 1929: “I did a very elaborate treatment of taxation and savings
which was cut out by Maynard; rightly as it was too involved in comparison
with the conclusions which were feeble” (Besomi 2003, 104, letter 158).
Therefore, Ramsey’s two published economics articles and the set of
notes published here taken together highlight important aspects of the
approach to economics of the young mathematician by métier who was
strongly drawn to economic science (Keynes [1933] 1972, 324).
First Set of Notes
IV
Let us now consider how to modify our equations in order to show the
effect of taxation ((on saving)), supposing that the public as a whole saves
as if it were discounting future utilities at a rate ρ, but otherwise keeping
its utility schedules unaltered. Suppose an income tax is levied at a con-
stant rate λ but is remitted on savings to a constant extent μ i.e. the total
tax paid is λ times consumption plus λ – μ times savings, then it is not
diffi cult to see that equations (2) and (3) become3
474 History of Political Economy 41:3 (2009)
4. Editor’s note: Ramsey (1928, 544) explains that equation (1) is the economy-wide resource
constraint. It states that savings, or investment, (dc(t) / dt) plus consumption (x(t)) equal income,
which is a function of labor (a(t)) and capital (c(t)): d—c(
dt
t—) + x(t) = f(a(t)), (c(t)).
5. Editor’s note: Unclear word.
6. Editor’s note: In the following equation, Ramsey went from the fi rst to the second line
by using the expressions for u(x) ⋅ ∂∂—f
a and u(x) ⋅
∂∂—f
c coming from (11) and (12), when ρ = 0. In the
fi nal line, he collected terms and used the fact that d—u(
dt
x—)—dt
dx = —du
dx.
v a = 1 –f
au x (11)
and
ddtu x =
1 – λ
1 – μ
∂ f
∂c– ρ u x (12)
The new form of equation (1) depends on what the government does
with the revenue it raises:4 it is convenient to consider two alternative
hypotheses. The fi rst is that the revenue is all transferred by the govern-
ment back to the public in the form of pensions, dividends on war loan etc
to be spent or saved by the public at its discretion. This income will then
be supposed5 liable to tax as is the case at present with war loan dividends
so that whether it is saved or spent will be affected by the rates of taxation
just as in the case of any other income. On this hypothesis equation (1)
will be unaltered, as on balance no income is taken from the public.
The second hypothesis is that the government raises money to spend
itself on purposes separate from those on which the public spends its
money, and that this government expenditure does not alter the utility
of private incomes. In this case we shall say that the revenue raised is
exhausted by government, and equation (1) must be replaced by
1 –dcdt+ x = 1 – f a,c (13)
Let us take fi rst the hypothesis that the revenue is all exhausted; then
if ρ = 0 we can easily obtain an analogue to (4): for we have6
ddx
u x · f a,c =dudxf a,c + u x
f
adadx
+ u xf
cdcdtdtdx
=dudxf a,c +
v a
1 –
dadx
–1 –
1 –
du x
dt
1 – f a,c – x
1 –
dtdx
=x
1 –
du
dx+
v a
1 –
da
dx
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 475
7. Editor’s note: Ramsey (1928, 544) denotes “by U(x) the total rate of utility of a rate of
consumption x, and by V(a) the total rate of disutility of a rate of labour a.”
8. Editor’s note: Ramsey (1928) assumes in section II that returns to capital, r, and the
returns to labor, p, are constant and independent in order to be able to (1) represent his solution
in a diagram, (2) extend it to the case of an individual who lives a fi nite time, and (3) extend
it to the case of time-discounting.
9. In this case μ might be greater than one of λ1, λ2 but not both, and the tax paid by a
particular person could then be negative; i.e. he would receive a bonus.
10. Editor’s note: In the equations to follow, as in the published paper (p. 550), y denotes the
“unearned” income (which is equal to consumption minus “earned” or labor income), and W(y)
and w(y) correspond to the total and marginal utility of unearned income, respectively. In the
second equation Ramsey used a version of (11) in which λ is replaced by λ1 and the fact that
∂f(a, c) / ∂a is now equal to p. The third equation comes from the defi nition of W(y) = ∫w(y)dy
and the fact that, from the fi rst equation, dy = dx (1 – λ1)p ⋅ da.
so that u(x) f (a, c)(1 – λ) = xu(x) – U(x) + V(a) + K or7
1 –dcdt= f a,c 1 – – x =
K – U x + V a
u x (14)
so that the onerous part of saving, as opposed to the proportion μ provided
by the exchequer, is related to marginal and total utility in the same way as
before; in the case of an eternal community K will as before be bliss, but
taxation may reduce its value as it reduces the maximum obtainable utility
if this is not equal to the maximum conceivable.
Further if we again make the assumption8
f (a, c) = pa + rc, ρ ≠ 0
we can conveniently allow for different rates of taxation λ1 and λ2 on
earned and unearned incomes respectively as well as a rate of remission
μ on savings.9 λ in equation (11) has then to be interpreted as λ1 and in
equation (12) as λ2 and instead of (13) we have
1 –dcdt+ x = 1 –
1pa + 1 –
2rc (15)
We now put10
y = x – 1 –1pa
w y = u x =v a
1 –1p
and
W(y) = U(x) – V(a)
(15) then gives 1 – dcdt = 1 –
2rc – y and (12) dwdt +
1 – λ2
1 – μr – ρ w = 0
Eliminating t and integrating for c in terms of w we get
476 History of Political Economy 41:3 (2009)
11. Editor’s note: In the text struck out by Ramsey one reads: ((The last question I shall dis-
cuss is the bearing of the equations developed in section IV on the problem of raising . . .)).
12. A Study in Public Finance p. 138.
13. See Economic Journal March 1927 p. 59.
1 – μ dcdt
=K – ∫w
σy dy
wσy
where σ =r 1 – λ 2
1 – λ 2 r – ρ 1 – μ .
Similarly if the revenue is all transferred, putting
y = x – pa
w y = u x
W y = U x –V a
1 –1
we get
dcdt
=K – ∫w
σy dy
wσy
where σ is now r 1 – μ
1 – λ 2 r – ρ 1 – μ .
V
In conclusion let us turn to the problem of raising a given revenue with
minimum sacrifi ce, distributional and administrative considerations
being disregarded.11
In the fi rst place, if λ1, λ2, and μ can all be varied independently the
problem is easy unless the revenue required is very large. The solution
is evidently given by λ1 = 0, λ2 = μ; i.e. there should be no taxation on
earned income, and the tax on unearned income should be offset exactly by
a bonus on saving. The revenue would in effect be raised entirely off capi-
tal existing at the time when tax is fi rst imposed, and the taxation would
have no “announcement effects,” to use Prof. Pigou’s phrase.
((Secondly)) But it may not be possible to tax earned and unearned
incomes at such different rates, so let us suppose instead that they are
taxed equally and inquire to what extent a uniform income tax should be
remitted on savings. Prof. Pigou in discussing this question12 refers to a
result obtained by me13 according to which uses of income for which the
demand is relatively elastic should be taxed at a lower rate than those for
which it is relatively inelastic; he concludes that as the demand for savings
is more elastic than that for consumption, as far as announcement aspects
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 477
14. Op. cit p. 136.
15. p. 59.
16. Editor’s note: The pages of Ramsey’s manuscript containing this paragraph are clearly
misnumbered. The text, as presented here, follows from page 37 to page 39 and then back to
page 38 and fi nally page 40.
go there should be differentiation in favour of savings. This means, or
should mean, that not merely should savings be altogether exempted from
income tax but that they should actually be rewarded with a bonus; for ((as
he has pointed out)) on Prof. Pigou’s view,14 unless exempted from income
tax they are really taxed twice, once when originally made and again when
they earn interest. I ((cannot, however, agree with)) do not dispute the sound-
ness of this conclusion, ((and)) but I ((am sure)) want to point out that ((my
result)) the result of mine to which he refers really cannot be used to sup-
port it. The reason for this is that in a problem covering a considerable term
of years saving cannot be considered simply as a use of income with its
own utility. Its utility is indirect and arises from the consumption it makes
possible later; it is therefore a part of the process of production rather than
of that of consumption, and it is evident that the ((kind of)) reasoning by
which I proved the result to which Prof. Pigou refers ((simpl[y])) cannot
be applied at all. The only case in which it can be applied is when we are
concerned with a very short period and a system of taxes which are only
to last for that short period and create no expectation of similar taxation in
future. In this case we should, as I showed in my paper,15 take the elasticity
of demand for saving as being infi nite.
Prof. Pigou’s solution must therefore be set aside and we must tackle the
problem by means of our equations. It is, however, ((unbearably)) extremely
complicated unless we assume that the supply of work is fi xed. In this
case if all the revenue is transferred back to the public there will be no
sacrifi ce at all provided μ = λ, since the two equations to determine x
and c are16
dcdt+ x = f a,c (1)
and
ddtu x = –
1 – λ
1 – μ
∂ f
∂c– ρ u x (12)
which are unaltered by raising λ, the rate of income-tax, to any extent
provided we always keep μ equal to it. In this case, therefore, income tax
478 History of Political Economy 41:3 (2009)
17. Editor’s note: This paragraph in the manuscript fi nished as “bonus, and this.”
18. B is ((unaffected by)) independent of λ and μ since on our present assumptions it is
the maximum conceivable rate of utility, which is bound to be obtainable or approachable in
the limit.
19. Editor’s note: In writing the exponential of the following equation Ramsey seems to want
to make a distinction between the return to capital (a cursive r) and the interest rate (a printed r)
used as the discount rate, as already mentioned. However, he used a cursive r to denote the inter-
est rate in the sentence preceding the equation. I make no such distinction here.
20. If it did not, some of our savings would be being left to earn compound interest for
ever, and so might as well never have been made.
should be completely remitted on savings, but these should receive no
further bonus.17
If, on the other hand, all the revenue is exhausted this conclusion is still
I think true provided that ρ = 0 and that, as we have often assumed before,
f (a, c) = pa + rc, but it ceases to be in the least obvious. All indeed that I
have been able to prove is that μ = λ gives a stationary value of the sacri-
fi ce not that it gives a true minimum. There seems, however, no reason to
doubt that the stationary value is in fact a minimum.
Equations (13) and (14) give us, if we absorb V(a) in B,18
1 –dcdt= 1 – pa + rc – x =
B – U x
u x (16)
The rate of revenue obtained at any time is given by
pa + rc – dcdt
and the total revenue obtained throughout time discounted at the rate of
interest r is given by19
R = pa + rc – dcdt
e–rtdt
0
=par + c
0+ –
dcdte–rtdt
0
since ce–rt vanishes at ∞.20
Now if we put 1 –1 –
= , we have
u x = u x0 e– rt, dtdx= –
u x
ru x
and therefore using (16)
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 479
R =pa + rc
0
r +–
1 – u1/x0
B – U x
u xu1/x dt
0 x0
(17)
=pa + rc
0
r +1 –
r · u1/x0
B – U x
u2x
u1/x – u x
x0
dx
Now the total amount, L, by which utility falls short of bliss is given by
L = B – U x dt0
=1
· r
B – U x
u x– u x dx
x0
(18)
We wish to show that the minimum value of L subject to (17) and to
1 – pa + rc0– x
0=B – U x
0
u x0
(19)
(from (16)) is given by λ = μ or κ = 1.
Differentiating (18) and putting dL = 0 we get
r · L · dk =B – U x
0
u x0
· u x0dx
0 (20)
Differentiating (19)
– pa + rc0d – dx
0= –
u x0
u x0
dx0–B – U x
0
u2
x0
u x0dx
0
from which using (20) we get
r ⋅ L ⋅ dκ = ( pa + rc0)u(x0)dλ (21)
If now the equation got by differentiating (17) and putting κ = 1 in the
result is an algebraic consequence of (20) and (21), it will follow that κ = 1
gives at least a stationary value of L.
But differentiating (17) and putting κ = 1 in the result we get
pa + rc0d –
d
u x0
B – U x
u x– u x
x0
dx = 0
480 History of Political Economy 41:3 (2009)
21. Editor’s note: On the back of this last page of the manuscript Ramsey wrote: “Memo:
correct III α to say equil may be attained not with r = ρ but with bliss.”
22. Editor’s note: Ramsey probably refers here to the national debt, with the following
equation being a government budget constraint.
or r ⋅ L ⋅ dκ = ( pa + rc0)u(x0)dλ which is identical with equation (21).
The required result is therefore established.
It is worth stating explicitly that we have assumed in this argument that
the rates of tax once imposed are never to be varied, and that by a given
revenue we mean a given supply of goods discounted to their present value
at the rate of interest, which is the rate at which the government could bor-
row or lend so as to obtain the use of its revenue in such quantities at such
times as it desired.21
Second Set of Notes:
“Savings Problem Abstract”
(1) No tax:
dc
dt+ x = f a,c (1)
v a =f
au x (2)
–d
dtlog u x =
f
c– (3)
(2) Tax revenue R rate λ with remission of μ on savings:
R = f a,c –dcdt
(0′)
v a = 1 – λ∂ f
∂au x (2′)
–ddtlog u x =
1 –
1 –
f
c– (3′)
if S of revenue exhausted
dcdt+ x = f a,c – S (1′)
[Side note:] esp. S = 0 or R
dcdt1 – + x = 1 – f
[end of side note]
if nat debt22 = B
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 481
23. Editor’s note: Unclear word.
24. Editor’s note: Unclear word; it looks like “reat” or “sent.”
25. Editor’s note: On the left side of the manuscript page with case (5), Ramsey wrote
vertically, in the middle of the page, “These equations can be solved geometrically.”
26. Editor’s note: Here Ramsey seems to make the already mentioned distinction between
the return to capital and the interest rate. However, in the following equations he seems to use
both interchangeably.
R1= f a,c + r
0B –
dcdt+dBdt
(0′′)
(3) conversion into money of constant marginal utility (money by dashes):23
if λ = μ = 0 c′ = u ⋅ c etc
if b = f – af
a– c
f
c= real24
dc
dt+ x = b + a + c
(4)
(4) Derived equation if ρ = 0; S = R:
1 –dc
dt=B – U x
u x+ V a (5)
{workout d—dx
(uf )}
[Side note:] λ on consumption r on saving = λ on both –μ on saving
∴ λ – μ = r, 1 – μ = 1 + r – λ [end of side note]
(5) Solution if f (a, c) = g(a) + rc and either S or R – S a known f[u]n[ction]
of a, x.25
Take S fi rst
v(a) = (1 – λ)g′(a) ⋅ u(x) (2′)
gives x as f[u]n[ction] of a or conversely let
y = x + S – g a
w y = u x
W y = w y dy = U x –V a
1 –if S = 0
Then dcdt= rc – y, dw
dt+
1 –
1 –r – w = 026
∴rc – y + dcdw
1 –
1 –r – w = 0
dc
dw+
r1 –
1 –r –
c
w=
y
w1 –
1 –r –
482 History of Political Economy 41:3 (2009)
Or
c ⋅ wr/
1 – λ
1 – μr – ρ
=y
1 – λ
1 – μr – ρ
w
r1 – λ
1 – μr – ρ
– 1
dy + B
=yr w
r/1 – λ
1 – μr – ρ
1r w
r/1 – λ
1 – μr – ρ
dy +– Br
Or
dcdt
= rc – y =B – ∫w
r/1 – λ
1 – μr – ρ
dy
wr/
1 – λ
1 – μr – ρ
y
(6)
also w y = Ae–
1 – λ
1 – μr – ρ t
Secondly if R – S = known f[u]n[ction] take
y = x – 1 – g a – R – S
w y = u x
W y = U x – V a if R – S = 0
1 –dcdt= 1 – rc – y dw
dt+
1 –
1 –r – w = 0
∴ 1 – rc – y + 1 –dcdw
1 –
1 –r – w = 0
dc
dw+
r 1 –
r 1 – – 1 –
c
w=
y
w 1 – r – 1 –
or 1 –dcdt= 1 – rc – y =
B – wr 1 – / 1 – r – 1 –
dy
wr 1 – / 1 – r – 1 –
y (7)
Special Case
a = const, S = R, f (a, c) = b + rc, u(x) = 1 / xα = (1 / Aα)e–σt, σ = r1 – λ
1 – μ
∴ x = Ae/ t
1 –dc
dt+ x = 1 – b + rc
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 483
dc
dt– c =
1 –
1 –b –
A
1 –e
/ t
∴ c = Be/ t
–1 – b
1 –+
A
1 – 1
1 –
e/ t
if to ∞ B = 0.
A = 1 –1
1 –c0+br
= 1 –1
1 – rc0+ b
[Side note:] α > 1 or no spending [end of side note]
Utility = Bliss – 1
– 1
1
x– 1
– 1 Loss of U =1
A– 1e–
– 1t
Loss0
=1
A– 1
·– 1
=1 –
– 1 r 1 –
·
– 1
– 1– 1
1 –– 1
rc0+ b
– 1
= K ·1 –
1 –
Revenue = b + rc –dc
dt
= ·rA
1 – 1 1 –e
/ t–1 – 1 –
e/ t
=A
– 1e
/ t
1 ––1 –
Discounted at rate r and integrated to ∞
=A
– 1 1 ––1 –
1
r –
= K1 –
–1 –
1 – 1 –
1 – – 1 –
= K ·1 – – 1 –
1 – – 1 –
484 History of Political Economy 41:3 (2009)
1 – μ = μ0, 1 – λ = λ0, loss = —λμ
0
—0
α
K · rev = c =1 –
0 0– 1 –
0 0
0–
0
< 1. Take small
∴ 0=
A01 – c
1 –0+
0– c
=01 – c
1 – c – – 10
1
loss0
– 1
1 – c – – 10
for min loss
– 1 1 – c – – 10= 0
∴ 0 = 1 – c, = c
u0 =1 – c
2
1 – c= 1 – c, = c
∴ =
∴ savings should be exempted entirely.
f (a, c) = pa + rc
λ0 = 1 – λ, μ0 = 1 – μ, σ = —μλ0
0
—r
If
u(x) = α – x = Ae–σt
v(a) = α – β = λ0 p(α – x)
= λ0 pAe–σt
a = β + λ0pAe–σt if R = S.
0
dcdt+ x =
0rc + p +
0p2
Aet
∴ dcdt– c = p 0
0
–
0
+
Ae– t
0
2
p2
+ 1
0
∴ c = a
0r–p
r–
A2
0p
0
2
p2
+ 1 e– t+ 0e
t
A = 2 ·0
–0p – 4
0rc0
0
2
p2
+ 1
c = c0+1
r·
0
– p – rc01 – e
– t
Const – U(x) = 1–2 (α – x)2 = 1–
2A2e–2σt
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 485
Loss throughout time = A2
4σ =μ 0
λ 0
⋅α – λ 0 rc 0 + pβ
2
1 + λ0
2 p 2 2
Revenue = 1 –0rc + pa – 1 –
0
dcdt
= 1 –0
rc0+
a
0
– p – rc0 1 – e– t
+ p + p2
0Ae
– t–
– 1 –0
0
0 0
– p
=1 –
0
0
0p +
1 –0
0
–1 –
0
0
+ 1 –0e– t·
·2p
2
0
p2
0
2
+ 1
· –0p –
0rc
0–
–0p –
0rc
0
0
=1 –
0
0
0p +
1
0
–1
0
+
+1 –
0
0
2p2
0
2
– 1
p2
0
2
+ 1
–0p –
0rc
0
u x = – x = Ae– t
, x = + Ae– t
v a = a – =0p – x
= 0 pAe– t, = 0
r
0
a = +0pAe
– t
debt D to be paid off (S = 0)
dcdt+ x = pa + rc
dcdt– rc = p +
0p2
+ 12
Ae– t–
c =– p
r–
A0
r0–
0
0p2
+ 1 e– t
∴ c = c0+
– p
rc01 – e
– t
A =– p – rc
0
1 +0p2
0+
0
0
Const – U x =1
2– x
2
=1
2A2
e–2 t
486 History of Political Economy 41:3 (2009)
[Side note]: If work fi xed, p2λ0 term omitted; then loss = f[u]n[ction] of λ 0
μ0;
min if this = 1 [end of side note]
∴ Loss =A2
4= 0
40
·0+
0
2
0
2
– p – rc0
2
1 + p2
0
2
=0+
0
2
0 01 + p
2
0
2·
– p – rc0
2
4
dDdt+ 1 –
0pa + rc + rD – 1 –
0
dcdt+dDdt
= rD
∴ 0
dDdt
–0rD = 1 –
0
dcdt
– 1 –0pa + rc
= 1 –0
0
0
– p – rc0– 1 –
0 0p2
A + 1 –0
– p e– t
– 1 –0p + rc
0
Loss of utility = B – U x + V a dt0
= B – U x + V a–u′ x
u x
1 – μ
1 – λ
1rdx
x0
X
=1r
1 – μ
1 – λ
B – U x
u xu′ x dx
x0
X
+ V av′ a
v ada
0
a0
⋅
Initial conditions
(1) S = 0.
rc0 – x0 + pa0 =B – ∫ x0– pa0
X
w1 – μ
1 – λ y dy
w1 – μ
1 – λ x0 – pa0
where w(y) = u(x) = v a
1 – p, dy = dx – pda
∴ rc0 + pa0 = x0 +
B – ∫ x0
X
u1 – μ
1 – λ x dx + 1
1 – λ p
⋅ ∫0
a 0
v a da
u x0
1 – μ
1 – λ
λ – μ
1 – λ
λ – μ
1 – λ
1 – μ
1 – λ
(2) S = R
1 – rc0+ pa
0= x
0+B – U x
0+ V a
0
u x0
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 487
Revenue
(1) S = 0 reckoned by marginal utility
Rev = pa + rc – dcdt
u x dt0
= 1r λ pa + rc –μ dc
dt–u′ x
1 – μ
1 – λdx
x0
X
=1r
1 – μ
1 – λλx + λ – μ dc
dtx0
∞
u′ x dx etc⋅
(2) S = R. Revenue reckoned at marginal utility
Rev = pa + rc – dcdt
u x dt0
=1 –
x · u x dt +1 –
1 –0
dc
dtu x dt
0
=1 –
1 –
1 – rx0u x0 + B – U x0 +
–
1 – 1 –loss
if a is fi xed
Rev =1 –
1 –
1
rrc
0+ pa
0u x
0+
–
1 – 1 –loss
Define: v =1 –
1 –
rRev = λv1 – λ
rc0 + pa0 u x0 + 1 – v1 – λ
B – Uu –u′ dx
x0
X
for min loss and given rev
0 = dv B – Uu · –u dx
x0
X
+ vB – U u
u dx0
+ d rc0+ pa
0=
B – U u
u2
dx0
0 = v
1 – λ2dλ + λ
1 – λdv rc0 + pa0 0u x0 + λ
1 – λv rc0 + pa0 u′dx +
+ 1 – v1 – λ
B – Uu
u′dx0 0+ dv1 – λ
+ 1 – v
1 – λ 2dλ B – U
uu′ dx
x0
X
⋅
488 History of Political Economy 41:3 (2009)
[Defi ne:] Z = ∫x0
XB – Uu –u′ dx
dλ dν dx0
ν B – Uu u′
rc0 + pa0
B – U u′
u 2
ν
1 –λ 2rc0 + pa0 u
+1 – ν
1 – λ2 Z
λ
1 – λrc0 + pa0 u
– Z
1 – λ
λ ν
1 – λrc0 + pa0 u′
+1 – ν
1 – λ
B – U u′
u
= 0
if λ = μ, v = 1
[Some additional equations which are hard to make sense of.]
Attempt to include variable rate of discount.
ddt
u x t = –f
cu x t
ddtlog u x t = –
f
c
ddtlog u x +
t
t= –
f
c
u x · t = Ae–rt
u x =Ae
–rt
t
w y =Ae
–rt
t= A t , y = t
y = rc – dcdt
ddt
w y · t = – r · w y · t
ddt
c · w y · t =dcdt
– rc w y t = –y · w y t
Duarte, ed. / Ramsey’s Notes on Saving and Taxation 489
27. Editor’s note: On the right-hand side of the following equation, Ramsey seems to have
mistakenly multiplied the term in square brackets by A. In this equation, I inserted the left-hand-
side term in square brackets.
[Side note:]27 ddt w y = w y
dy
dt=A – rw y –
t
tw y [end of side note]
dc
dt– rc = t
c = Ae–rt
+ ert
e–rt
t dt0 t
t
dcdt= re
rte–rt
t dt0 t
t
= rert
t e–rtdy
t
dcdt
= r χ t e–rtdt
0
∞
= e–rtdt
0
∞
= rw′ y
Aw y –r –ϕ′ t
ϕ t
dy
t
∞
References
Besomi, D., ed. 2003. The Collected Interwar Papers and Correspondence of Roy
Harrod. Vol. 1, Correspondence, 1919–35. Cheltenham: Edward Elgar.
Duarte, P. G. 2008. Another Chapter in the History of Ramsey’s Optimal Feasible
Taxation. Working paper.
Keynes, J. M. [1933] 1972. Essays in Biography. Vol. 10 of The Collected Writings of
John Maynard Keynes, edited by Elizabeth Johnson and D. E. Moggridge. London:
Macmillan.
Pigou, A. C. 1928. A Study in Public Finance. London: Macmillan.
Ramsey, F. P. 1927. A Contribution to the Theory of Taxation. Economic Journal
37.145:47–61.
———. 1928. A Mathematical Theory of Saving. Economic Journal 38.152:543–59.
top related